1 validation of wrf surface meteorology simulations robert fovell university of california, los...
TRANSCRIPT
![Page 1: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/1.jpg)
1
Validation of WRF surface meteorology simulations
Robert FovellUniversity of California, Los Angeles
![Page 2: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/2.jpg)
2
Background
• Ph.D., 1988 - University of Illinois (Atmospheric Sciences)• Postdoc at University of Washington (Atmospheric
Sciences)• 1991 - joined faculty at UCLA (Atmospheric & Oceanic
Sciences)• Full Professor, Vice Chair, Undergraduate Faculty Advisor• Chair of the Faculty of the College of Letters & Science• UCLA Distinguished Teaching Award• Chair of Mesoscale Processes Committee (American
Meteorological Society)• Member, WRF Ensemble Working Group
![Page 3: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/3.jpg)
3
Selected research on squall-lines
Fovell and Ogura (1988, 1989)storm maintenance, viability,
microphysics influence on structure
Fovell, Durran and Holton (1992)stratospheric gravity waves
Fovell (2002)Fovell, Mullendore and Kim (2006)
upstream influence and cell initiation
Fovell and Tan (1998, 2000)storm unsteadiness
![Page 4: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/4.jpg)
4
Sea-breeze and convective initiation
Dailey and Fovell (1999)Why convective rolls modulate
sea-breeze convection
Fovell and Dailey (2001) Fovell (2005)
Convective initiation aheadof the sea-breeze front
![Page 5: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/5.jpg)
5
Hurricane tracks and microphysics
Hurricane Ike - (a) WRF convection ensemble; (b) NHC multi-model ensemble; (c) GFS ensemble
Fovell and Su (2007)
(a)(b)
(c)
![Page 6: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/6.jpg)
6
Hurricane tracks and microphysics
Fovell, Corbosiero and Kuo (2009)
![Page 7: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/7.jpg)
7
A few other papers
• Fovell and Fovell (1993)– Very early application of cluster analysis to
geophysical data
• Berk and Fovell (1999)– “willingness to pay” for climate change avoidance
• Hughes, Hall and Fovell (2006)– Effect of topography on diurnal cycle in Los
Angeles
• Hughes, Hall and Fovell (2008)– Los Angeles rainfall and topographic blocking
![Page 8: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/8.jpg)
8
Outline
• Summer event (August 2000)
• Winter event (January 2001)
![Page 9: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/9.jpg)
9
Summer event
• August, 2000 high-ozone event– August 1st-3rd– Very “asynoptic” situation
• WRF and MM5 models– 3 telescoping domains– 4 km resolution over central CA– Emphasize WRF owing to ARB dissatisfaction w/ MM5
• Comparison of model with available surface observations
• “Achilles heel” of regional-scale modeling
![Page 10: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/10.jpg)
10
1 August 2000 @ 12Z(5 AM PDT)
87˚F
![Page 11: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/11.jpg)
11
WRF/MM5 domain setup
36, 12 and 4 km
![Page 12: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/12.jpg)
12
Observations for comparison
• Mesowest archive (Univ. of Utah)• California ASOS and RAWS stations
– ASOS - mainly airports, populated areas– RAWS - many in remote areas, especially
higher elevation
• 2 m temperature, dew point and 10 m wind speed– Need to get the mesoscale circulation right
![Page 13: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/13.jpg)
13
Model vs. station elevation
76% within ±200 m
![Page 14: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/14.jpg)
14
ASOS+RAWS “representative stations”
(WRF elevation error under 200 m; < 5 missing hours)
156 stations
![Page 15: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/15.jpg)
15
Results using WRF version 2
• Eta and FNL initializations
• Investigated surface physics, radiation, model start time, simulation length, nudging
![Page 16: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/16.jpg)
16
Experimental design
Control runs:Simulation set with varying physics combinations
Starts at 00Z (5 PM PDT July 31)
![Page 17: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/17.jpg)
17
Experimental design
Delayed start runs:Tests whether better to start with more stable conditions
![Page 18: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/18.jpg)
18
Experimental design
Day 2 (and Day 3) runs:Tests whether model behaves similarly on a
similar (Day 2) and cooler (Day 3) day
![Page 19: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/19.jpg)
19
Experimental design
Longer runs: Does model need spin-up time?Nudged runs:
Tests whether model performs better when prodded
![Page 20: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/20.jpg)
20
August 1 temperature(representative stations)
• Underpredicted 2m T in afternoon
• Did no better with– Starting at 12Z instead– Using FNL– August 2nd, starting 00Z
on 2nd– August 2nd, starting 00Z
on 1st– August 3rd, a cooler day– Nudging WRF– Using MM5
Simulation started00Z 1 August 2000Using Eta analyses
![Page 21: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/21.jpg)
21
August 1 dew point(representative stations)
TD
Noah
• Noah surface scheme better than TD
• Better Td predictions marginally improved T forecast skill
![Page 22: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/22.jpg)
22
August 1 wind speed(representative stations)
• Simulations too “windy” overnight but not to blame for afternoon cold bias
• WRF superior to MM5
• Wind errors largest when overall model doing best
MM5
WRF
![Page 23: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/23.jpg)
23
Errors among different simulations highly correlated
^ diff. radiation - diff surface v ^ diff. rad & sfc - diff models v
![Page 24: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/24.jpg)
24
T and Td errors
< Td errors w/different radiation scheme
T and Td errors >not correlated
(not correlated w/ elevation error either)
![Page 25: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/25.jpg)
25
Simulations with WRF v.3
• modified TD scheme• improved nudging
• slope radiation & topo shading• extend study to include NARR,
PBL variations, landuse contributions
![Page 26: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/26.jpg)
26
Avg. T for representative stations
xX = relatively large #of data dropouts
![Page 27: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/27.jpg)
27
Avg. T for representative stations
Added simulations initialized with Eta and FNL,Noah and TD surface, various PBLs.
Results very similar to WRF v.2.
![Page 28: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/28.jpg)
28
Avg. T for representative stations
Added simulations initialized with NARR.(NARR actually has smallest error before sunrise.)
NARR runs;
2 different PBL schemes
![Page 29: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/29.jpg)
29
Slope radiation and shading
Hypothesis: sets up T gradient acrossCentral Valley, inducing mesoscale circulation
![Page 30: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/30.jpg)
30
Slope radiation and shading
![Page 31: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/31.jpg)
31
Source of error:bottom-up vs. top-down?
Asynoptic, low wind conditions: expect errors strongly local &
bottom-up
![Page 32: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/32.jpg)
32
Landuse type distributionamong representative stations
landuse cat. type count
1 urban 10
3 Irrigated
cropland
5
7 grassland 5
8 shrubland 32
9 shrub/grass 6
10 savanna 39
11 deciduous forest 6
14 evergreen forest 52
16 water 2
Cat. 8, 10, 1479% of total
![Page 33: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/33.jpg)
33
T errors by landuse categoryfor representative stations
Categories 8, 10 and 14 account for 79% of stationsCategories 8 (shrub) and 14 (evergreen) reveal cold biases
[and represent 54% of all stations]
![Page 34: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/34.jpg)
34
T errors by landuse categoryfor representative stations
Experiment to manipulate surface/soil characteristicsfailed to resolve temperature bias
![Page 35: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/35.jpg)
35
Avg. T for representative stations
Added Eta-nudged run.Nudged free atmosphere only. Nudging PBL also
didn’t help. Also nudged nests.
nudged run
![Page 36: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/36.jpg)
36Colors and digits: nudged-unnudged
![Page 37: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/37.jpg)
37
^ Delta ‘sea-breeze’’weaker
![Page 38: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/38.jpg)
38
Avg. T for representative stations
Nudging nests experiment:Nudging all best, nudging just D1 or D1&D2
nearly indistinguisable
nudged all
nudged D1
nudged D1 & D2
![Page 39: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/39.jpg)
39
Nudging NARR?
Nudged EtaUnnudged Eta
Unnudged NARR
![Page 40: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/40.jpg)
40
Nudging NARR?
Nudged EtaUnnudged Eta
Unnudged NARR
Nudged NARR
Difference in nudged vs. unnudged NARRstarts right after sunrise. Why?
![Page 41: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/41.jpg)
41
∆T nudged-unnudged for NARR runsCentral Valley, near Kettleman City
time
![Page 42: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/42.jpg)
42
NARR-Eta hybrid initialization
• Feature of WPS - combine data sources
• Tested NARR 3D information with surface/soil initialization from Eta
• Result: little different from unnudged NARR-only run (i.e., still very poor after sunrise)
![Page 43: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/43.jpg)
43
Comments
• MM5 and WRF both have difficulty simulating an asynoptic, heat-wave situation– Particularly acute for NARR initialization– Varying PBL, other physics, landuse, fails to cure
• WRF v.3 nudging produces highest fidelity results– Even for NARR initialization… bugs?– Unnudged skill loss apparently ‘top down’ from free
atmosphere
• LW radiation (in RRTM) bugfix released 10/23/08– Little impact on results (actually a little worse)
• Brute force (nudging) doesn’t cure/excuse bad physics (or bugs)
![Page 44: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/44.jpg)
44
Generalized results(subject to change as matrix gets filled out, more cases examined)
• Nudged runs better than unnudged• Nudging all domains better than outer alone• FNL better than Eta, much better than NARR• Noah better than TD surface scheme• YSU better than MJY PBL• Slope/shading effects unimportant• Local landuse variations apparently unhelpful• RRTM radiation bug fixes don’t help
![Page 45: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/45.jpg)
45
Winter case
January, 2001
![Page 46: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/46.jpg)
46
January 2 weather summary
• High pressure interior west– Santa Ana conditions in So. California– Light winds in Central Valley
• Persistent low fog in Central Valley
• Slow-moving N Pacific storm approaching
![Page 47: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/47.jpg)
47
12Z January 2 observations
^ IR
Surface map >
![Page 48: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/48.jpg)
48
24 h max RH on January 2nd
![Page 49: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/49.jpg)
49
WRF-ARW configuration
• Same three domains as August 2000 case• FNL data input• RRTM (pre-fix), Noah schemes• Microphysics (both 6 class):
– Seifert-Behing two-moment microphysics– Thompson scheme (single moment cloud)
• Kain-Fritsch cumulus parameterization in outer 2 domains
• No nudging
![Page 50: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/50.jpg)
50
Clear air turbulence case Storm anvil cloud shown (Fovell et al. 2008)
< Seifert-Beheng microphysics
< WRF Single Moment 6 class (WSM6) microphysics
Seifert anvil more sharply defined, comparable to obs
![Page 51: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/51.jpg)
51
Jan 2 near-sfc RH - Seifert
Blue = cloud water > 0.1 g/kg
![Page 52: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/52.jpg)
52
14Z January 2 (14 h forecast)
Seifert Thompson
Little difference at this time
![Page 53: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/53.jpg)
53
19Z January 2 (19 h forecast)
Seifert Thompson
Seifert fog breaks up more slowly
![Page 54: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/54.jpg)
54
21Z January 2 (21 h forecast)
Seifert Thompson
No fog remains in Thompson simulation
![Page 55: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/55.jpg)
55
Sacramento relative humidity
Obs
![Page 56: 1 Validation of WRF surface meteorology simulations Robert Fovell University of California, Los Angeles rfovell@ucla.edu](https://reader035.vdocuments.site/reader035/viewer/2022062722/56649f345503460f94c517a4/html5/thumbnails/56.jpg)
56
end